Ischemic stroke patients attributable to atrial fibrillation (AF) and rheumatic heart disease (RHD) are increasing in China, with a high rate of death and recurrence (1, 2). Although antithrombotic therapy (antiplatelets or oral anticoagulants) has been the mainstay for primary and secondary prevention of ischemic stroke in patients with AF and/or RHD, its use in clinical practice is suboptimal due to the feared risk of catastrophic intracerebral hemorrhage (ICH) (3). Therefore, risk stratification strategy is needed to balance the risk of ICH vs. ischemic stroke and guide the safe use of antithrombotic agents in Chinese stroke patients with AF/RHD, who have a heightened risk of ICH (4).

Cerebral microbleeds (CMBs) are key neuroimaging markers of hemorrhagic small vessel disease in the brain, which can be detected on bleeding sensitive magnetic resonance imaging (MRI) sequences such as gradient-echo T2^*-weighted imaging (T2^*-GRE) and susceptibility-weighted imaging (SWI) (5). Histopathologically, CMBs are perivascular hemosiderin deposits after small hemorrhages, indicative of bleeding-prone microangiopathy (6). They are commonly detected in healthy elderly and more prevalent in patients with ischemic or hemorrhagic strokes (7–9). The presence of CMBs has been reported to increase the risk of ICH up to 8-fold in a pooled cohort of ischemic stroke or transient ischemic attack (TIA) (10). This is more clinically relevant in AF/RHD patients with anticoagulation needs (11). Prior studies showed that CMBs appear to be more prevalent in patients with AF (12) and ≥5 CMBs powerfully predict future ICH risk in ischemic stroke patients with AF taking oral anticoagulants (13). Consequently, a challenging clinical dilemma emerged as whether antithrombotic decisions would be influenced by the presence of CMBs in patients with AF and/or RHD.

Since CMBs may represent hemorrhage-prone pathological states, the impact of different antithrombotic therapy on the presence of CMBs has been of great concern. Some previous studies showed that antiplatelet or anticoagulant use was significantly associated with the presence of CMBs in ischemic stroke patients (12, 14). However, the results regarding this question remained controversial and uncertain (15, 16). In addition, few studies (12, 17) on this question focused on stroke patients with AF and/or RHD. Besides, most studies observed CMBs on T2^*-GRE, not on SWI, though SWI has been proven to be more sensitive at detecting CMBs (18).

Hence, we aimed to investigate the relationship between prior antithrombotic therapy and prevalence of CMBs observed on SWI in ischemic stroke patients with AF and/or RHD. Furthermore, we analyzed this association in different locations of CMBs with the presumption that strictly lobar or deep/infratentorial CMBs may have different underlying pathogenesis and different bleeding risk (5).

In the study period, we enrolled 160 consecutive ischemic stroke patients with AF and/or RHD within 7 days of stroke onset at admission (West China Hospital, Sichuan University, n = 125; People's Hospital of Deyang City, n = 35). Of the included patients, there were 68 (42.5%) men and the median age at stroke onset was 71 years (age range 59–78 years). Thirty-eight patients (23.8%) had been treated with antiplatelets exclusively before admission (30 for aspirin, 4 for clopidogrel, 4 for both), 18 (11.3%) patients had been treated with warfarin exclusively, and another 10 patients switching between antiplatelets and warfarin were analyzed separately. CMBs were detected in 90 patients (56.3%). Of these, 36 had single CMBs, and 54 had multiple CMBs (≥2). CMBs were most commonly present as deep or infratentorial bleeding (53/90, 58.9%), followed by strictly lobar bleeding (37/90, 41.1%).

Table 1 shows the demographic and clinical characteristics of included patients with and without CMBs. Those patients with CMBs were more likely to have a history of hypertension (48.9 vs. 28.6%; P = 0.009), lower NIHSS score on admission (median 5 vs. 9; P = 0.001), more burden of WMH (75.6 vs. 55.7%; P = 0.008) and were on exclusive antiplatelet treatment three times (33.3 vs. 11.4%; P = 0.001) as commonly compared with those without CMBs. However, patients were comparable in terms of prior use of anticoagulants and coagulation function test.

Table 2 describes the baseline demographics and CMB distribution stratified by antiplatelet or anticoagulant use. Compared with non-antithrombotic users, those with prior antiplatelet use were older, more frequently to have CMBs, and had marginally higher prevalence of strictly lobar CMBs. Patients with warfarin had higher INR values but no significant difference was found in the presence or location of CMBs.

In a multivariable logistic regression model after adjustment for age, gender, history of hypertension, history of stroke or TIA, NIHSS score, INR and presence of WMH, the exclusive use of antiplatelet agents was independently associated with the presence of CMBs (OR 3.075, 95% CI 1.175–8.045, P = 0.022; Table 3). In terms of CMBs' location, prior antiplatelet use was associated with strictly lobar CMBs (OR 2.635, 95% CI 1.050–6.612, P = 0.039; Table 3) but not with deep or infratentorial CMBs. Specifically, in a subgroup of exclusive aspirin users, the significant results remained. Prior use of aspirin was still independently related with the presence of CMBs (OR 3.095, 95% CI 1.085–8.832, P = 0.035) and strictly lobar CMBs (OR 3.975, 95% CI 1.506–10.495, P = 0.005) after controlling the above confounders. However, no significant association was found between the exclusive warfarin use and the presence or location of CMBs either in the univariate or multivariate analysis.

In this study, we found that CMBs were highly prevalent in Chinese ischemic stroke patients with AF and/or RHD. Prior use of the antiplatelet agents (especially aspirin) was significantly associated with presence of CMBs and especially the strictly lobar CMBs, whereas the prior use of warfarin seemed to be not related to CMBs in any location.

The frequency of CMBs in stroke patients with AF/RHD was 56.3% in the present study, which fell broadly within the range of 18% to 68% in patients with ischemic stroke (23). When focusing on stroke patients with AF, the prevalence of CMBs was reported to be around 30% (11, 12, 24, 25). The higher prevalence of CMBs in our study might be related to ethnic differences or different imaging protocol. We carefully used the SWI sequence which has better spatial resolution and post-processing ability in detecting CMBs than conventional T2^*-GRE sequence (18).

With the high prevalence of CMBs and a possible link to bleeding tendency, there is growing interest in demonstrating the role of CMBs in AF patients who are frequently using antithrombotic agents. Our study showed that antiplatelet use was associated with presence of CMBs in ischemic stroke patients with AF/RHD. Although controversies existed in the literature (14, 26–28), our finding is consistent with three recent meta-analyses, which all showed similar associations in patients with ischemic stroke or TIA (15, 29, 30). Some of the previous studies suggested that different type of antiplatelets may have different effects on development of CMBs. The Rotterdam Scan Study showed that aspirin was associated with lobar CMBs while clopidogrel associated with deep CMBs in general population (31, 32). A Japanese study among ICH patients reported that only aspirin, but not clopidogrel, cilostazol, or ticlopidine, was associated with the presence of CMBs (27). We also found that prior use of aspirin was associated with the presence of CMB; however, we did not evaluate the impact of clopidogrel on CMBs due to the limited number of clopidogrel users. These discrepancies among studies may be due to the difference of study population and prevalence of CMBs. Furthermore, in terms of the duration of antiplatelet therapy, a Chinese study found that CMBs are more prevalent among aspirin users with longer duration (>5 years) than those receiving aspirin <5 years (14), whereas another study from Japan failed to show this association between long duration (≥10 years) of aspirin use and CMB presence in multivariable analysis after adjusting for hypertension and other confounders (16). It is important to note that most of these studies including ours are major cross-sectional design, a conclusion on the causal relationship between antiplatelet use and development of CMBs cannot be drawn.

Different topographical associations of CMBs with the prior antiplatelet therapy were detected in our study. Prior use of antiplatelets and especially aspirin were only significantly associated with occurrence of strictly lobar CMBs rather than deep or infratentorial CMBs. Our findings are in line with the Rotterdam Scan Study (31) and the newly published meta-analysis (30). This result may be explained by the different underlying pathophysiologic mechanisms of CMBs in different locations. CMBs located in strictly lobar brain regions are associated with cerebral amyloid angiopathy, while those located in deep or infratentorial brain regions are related to hypertensive vasculopathy (5). Intriguingly, some previous study reported that the aspirin-associated hemorrhages predominantly occurred in the lobar areas (33), suggesting that aspirin favors rupture of small vessels with cerebral amyloid angiopathy. In this aspect, it is speculated that antiplatelet use would be more related to risk of lobar CMBs other than deep or infratentorial CMBs. However, due to the small sample size in subgroup analysis, there might be somewhat overfitting and possibility of chance. Moreover, whether lobar CMBs are truly attributable to CAA in patients with AF is yet to be ascertained. Further pathological studies are needed to clarify this point and explore other conditions that modulate the link between antiplatelet use and lobar CMBs.

In our study, we could not confirm an association between anticoagulation and CMBs. This result should be interpreted cautiously due to the small sample size of warfarin users. Nevertheless, previous studies have provided conflicting results regarding this question. In the Rotterdam Scan Study (31) and a Korean study of asymptomatic elderly (34), no association was found between warfarin and presence of CMBs. Another Turkish case-control study also demonstrated that warfarin treatment had no effect on CMBs in ICH patients (35). Moreover, in a longitudinal study with ischemic stroke patients on warfarin therapy for 2 years, the mean duration of warfarin treatment was not significantly related to new development of CMBs (36). Conversely, there is also evidence linking warfarin use to CMB presence. A German retrospective study of 785 patients with ischemic stroke and TIA (16.3% with AF) showed that patients with prior use of anticoagulant agents were more likely to have CMBs as compared to those without; these CMBs tended to be in lobar location (12). Unfortunately, this association was no longer present after full adjustment for demographic and clinical variables in the multivariable analysis, with the age remained the only independent factor for CMBs (12). In a large population-based study in Netherlands, higher INR values were associated with CMBs; however, we did not find the association in our study (37), which might be partly due to few patients with targeted INR. Considering INR values representing the intensity of anticoagulation that may influence the occurrence of CMBs, further prospective studies are required to clarify this association.

Notably, we also found that hypertension is an important risk factor of CMBs, which was in line with previous studies (16, 38). The chronic hypertension and associated arterial stiffness have been reported to be linked to the presence of CMBs (39) and further contribute to higher risk of hemorrhagic stroke especially after possible acute treatment such as thrombolysis (40, 41). Therefore, it is reasonable to speculate that hypertension may explained at least in part the occurrence of CMBs in addition to the effects of antithrombotic therapy. Under impaired hemostasis by antithrombotic agents, the red blood cells are prone to extravasate through the vulnerable microvasculature and this process might be aggravated by the presence of hypertension. These results suggest that vascular risk factors should be carefully monitored when prescribing antithrombotic drug treatment to patients with multiple CMBs.

Studying the relationship between CMBs, antithrombotic agents and AF/RHD is of great importance. Firstly, patients with AF have a significantly higher prevalence of CMBs and they tend to receive anticoagulation and antiplatelets more frequently than patients without AF (24). Secondly, CMBs are small blood extravasations and these asymptomatic leaks could evolve into a symptomatic ICH when hemostasis is damaged under antithrombotic therapy (42). There is now clear evidence that ICH incidence is higher in patients with CMBs than those without CMBs in antithrombotic users with AF (43), and the risk of ICH associated with CMBs is higher in Asians (10). With the aging population and growing prevalence of AF/RHD patients in China, it is essential to investigate the correlation of CMBs with antithrombotic therapy and determine how this small vessel disease marker can help analyze the benefit-risk ratio for stroke prevention strategies.

Our study has several limitations. First, this is a hospital-based study with a relatively small sample size and low number of the patients treated with oral anticoagulants. Second, subgroup analysis of different type of antiplatelets was not performed because most of our patients used aspirin. Nevertheless, the significant association of antiplatelets as whole remained in the subgroup of aspirin users. Third, the durations of drug use were not recorded due to the incomplete data in the medical records. It is possible that the durations of antithrombotic agents influence the occurrence of CMBs. Fourth, the selection bias cannot be avoided because only part of patients underwent SWI sequence due to the disease severity and cost of multiple MRI sequences. Finally, given the cross-sectional and retrospective nature of the study, it is possible that some CMBs may have occurred before use of antithrombotic agents. Therefore, further prospective investigation is required to confirm these findings.

In conclusion, prior antiplatelet use is independently associated with the presence of CMBs in ischemic patients with AF and/or RHD, especially among those with lobar CMBs. Whether anticoagulants by itself could cause CMBs remains unclear. These findings warrant future prospective longitudinal studies to clarify the relationship among antithrombotic therapy, cerebral microbleed and macrohemorrhage in stroke patients with AF and/or RHD.

ML designed the study. YC, JfL, JieL, CW, and JinL collected the data. YC and JfL performed imaging analysis. YC and JfL performed statistics analysis. YC and JfL drafted the main part of the manuscript. BW and SZ provided support for imaging analysis. SZ, DW, and YW helped design the study and revised the manuscript. All authors approved the final version submitted for publication.